Table 2 Summary of measurement properties of the FSOC-S12, its interpretation from the initial analyses
From: Measuring a family sense of coherence: a rasch-based study extending dyadic data analyses
Measurement property | Interpretation and criteria | FSOC-S12: DIP Initial analyses | FSOC-S12: DIP Analyses after collapsed thresholds | FSOC-S12: SIP Initial analyses | FSOC-S12: SIP Analyses after collapsed thresholds | ||
|---|---|---|---|---|---|---|---|
Sample to item targeting | The relative distributions of item-threshold:person measures. | Mean person location close to zero [29]. Range of person locations Range of item-threshold locations | 1.15 logits (SD 1.02) -0.76-5.98 -0.50-0.79 | 1.13 logits (SD 1.13) -1.81-6.02 -0.95-1.16 | 1.51 logits (SD 1.30) -0.85-5.89 -0.55-0.93 | 1.34 logits (SD 1.45) -2.52-5.83 -0.98-1.28 | |
Response category function | Ratings should be consistent with the estimate of the underlying construct | Monotonically increasing thresholds [31]. | 4 of 24 had monotonically increasing thresholds | All had monotonically increasing thresholds | 1 of 12 had monotonically increasing thresholds | All had monotonically increasing thresholds | |
Item fit to the model | How observed data accord with the Rasch model. | MNSQ is recommended to be within 0.5-1.5 and ZSTD is recommended not to exceed ±2SD [32]. | MNSQ OUTFIT | 23 of 24 in recommended range | 23 of 24 in recommended range | All in recommended range | All in recommended range |
ZSTD OUTFIT | 21 of 24 inside recommended range | 22 of 24 inside recommended range | 9 of 12 inside recommended range | 11 of 12 inside recommended range | |||
MNSQ INFIT | All in recommended range | All in recommended range | All in recommended range | All in recommended range | |||
ZSTD INFIT | 22 of 24 inside recommended range | 23 of 24 inside recommended range | 11 of 12 inside recommended range | 11 of 12 inside recommended range | |||
Unidimensionality | Whether items appear to represent one common variable | Eigenvalue of unexplained variance in the 1st contrast of a PCA of fit residuals should not be greater than 2 [33]. | Eigenvalue | 3.44 | 3.60 | 1.71 | 1.75 |
Disattenuated Pearson correlations of the person measures close to 1 | Disattenuated Pearson correlation | 0.61 | 0.62 | 1.00 | 1.00 | ||
Local independence (LD) | There should be no dependence between two items, i.e., item responses should be independent of each other | Relative correlations of residuals should not exceed the relative Q3* cut off [45]: - 0.24 for dyadic-informant perspective 0.12 for single-informant perspective | LD for 9 of 276 correlations | LD for 9 of 276 correlations | LD for 3 of 66 correlations | LD for 4 of 66 correlations | |
Differential item function (DIF) | DIF occurs when items behave different for differently sub-groups, in this paper, between patients and their family members from the single-informant perspective | There should be no statistical differences between item estimates, p-value should not be significant be below 0.05 | DIF probability | N/A | N/A | Item 4: 0.01 Item 12: 0.01 | Item 4: 0.02 Item 6: 0.03 Item 8: 0.03 Item 12: 0.01 |
DIF ≥ 0.64 logits = moderate to large; DIF ≥ 0.43 = slight to moderate. | DIF size | N/A | N/A | Item 4: 0.36 Item 12: 0.32 | Item 4: 0.42 Item 6: 0.33 Item 8: 0.51 Item 12: 0.42 | ||
Reliability | The proportion of variance that is true variance | 0 implies all error and 1 implies no error | Person reliability | 0.87 | 0.89 | 0.82 | 0.84 |
Item reliability | 0.95 | 0.97 | 0.98 | 0.99 | |||